Mixed Music Recommendation System Research Based On Excellent Reviews

After years of development,the Internet has reached a stage of great development and prosperity.Various forms of data,such as text,pictures,Numbers,video,voice and so on,will be recorded and spread in the sea of the Internet.The development of the Internet has also brought about the problem of information overload.How to screen out the information that users are interested in from the sea of the complex network has been a hot topic of research.With the development of big data and artificial intelligence in recent years,recommendation systems are increasingly applied in all walks of life.By analyzing a large number of users’ historical records and behavior data,the recommendation system can find out the items that users may be interested in and recommend them to customers,or find people with similar interests,and then recommend the items that these people like to users.After continuous exploration and research by predecessors,the recommendation system has developed into a mature one.There are mainly user/item-based collaborative filtering technologies,model-based collaborative filtering,and content or knowledge-based recommendations,as well as various hybrid mode recommendation techniques.The recommendation system is also used in music.Music has become an indispensable entertainment element in modern biochemistry.Many people’s interests and hobbies include listening to music.You may not be able to sing,but you will certainly listen to it.There are many music applications in the market,each of which has built its own unique and perfect recommendation system.One of the most famous ones is Pandora Radio.The core of its recommendation system is a project called "music genome project".The music genome project aims to "capture the essence of music with the most basic level".Many musicians and music lovers manually screen more than 450 attributes to describe each song,and use complex mathematical algorithms to organize them and recommend songs with similar interests to users.The early music recommendation system started from music itself and marked the melody,rhythm and Musical Instruments of a song to find out similar music and recommend it to users.Later,recommendations based on social information were developed,mainly recommending music to users and others with similar interests.People always like reading the wonderful comments of many songs and sometimes they can understand the story behind the music.Why not use the wonderful comments under each song to mine relevant information to make a recommendation system? This paper will study the recommendation algorithm based on the traditional recommendation algorithm and add the social and cultural elements extracted from the wonderful comments under each music.The idea of the algorithm is inspired by singular value decomposition.Thinking reversely,this paper firstly analyzes the main keywords from each song as a "potential factor" through a certain analysis,and calculates the user-potential factor score matrix,song-potential factor score matrix,and finally combines with the user’s behavior data.A recommendation is made on the basis of the user-song scoring matrix.In this paper,on the basis of studying a large number of previous literatures,the commonly used recommendation algorithms and techniques are summarized,and a hybrid recommendation system based on brilliant reviews is designed according to the initial assumption.In the first stage,the potential factor algorithm was used to calculate the score matrix of the user’s songs.By comparing with the original score matrix,the first recommendation was made for the songs that were rated but not listened to by the user.On this basis,the collaborative filtering in the second stage was carried out to find out the users with similar interests and make the second recommendation.After the empirical analysis of the designed recommendation system,this paper used the selenium module of Python to crawl the data.After the sorting and data preprocessing,a total of 3,302 songs,33,020 reviewers and their comments,1,947 users and the total number of songs they listened to,and the top ten of the songs they listened to,were screened out and 941 potential factor labels were selected.In this paper,Python and SQL were used to further extract all non-zero scores from the user-song score matrix,and the data were sorted into the form of {user,music,rating},with a total of 393,717 pieces of data obtained.Finally,based on this data set,the mixed music recommendation system based on excellent comments is evaluated.The experimental design was to randomly divide the 393,717 pieces of data into 10 pieces,with 3 pieces as the test set and the remaining 7 pieces as the training set.In order to avoid overfitting of the evaluation results,we conducted several experiments to take the average value.Finally,the four evaluation indexes of accuracy,recall rate,coverage rate and novelty are obtained.The results show that the mixed music recommendation algorithm based on excellent comments has high recommendation accuracy.Finally,this paper summarizes the research methods and process,and puts forward three points that need to be studied further more.that is,to further improve and optimize the analysis of wonderful reviews and the extraction of potential factors;The user-item scoring matrix obtained by potential factors needs to organize a customer satisfaction survey.In the process of designing recommendation system,we should have a holistic view and not only focus on a certain module.